• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2008년도 하계종합학술대회
• /
• pp.1017-1018
• /
• 2008

In this paper, we propose a method that makes use of neighborhood relationship in 2D spectrogram of separated sources toward the generalization of the binary mask in Degenerate Unmixing Estimation Technique (DUET). A new generalized mask can be consist of five to ten mask. According to the new mask, the original power of the spectrogram in each frequency-time point is assigned. The result showed a smooth and tender wave-form, indicating a high speech separation performance compared to the original method.

• 전기학회논문지
• /
• 제60권12호
• /
• pp.2374-2379
• /
• 2011

The fundamental frequency of the voiced speech is estimated using the instantaneous frequency based on the symmetric higher order differential energy operator. The instantaneous frequency based on the symmetric higher order energy operator shows better frequency estimation result since it is aligned to the time instance of the signal. The speech is pre-processed by a lowpass filter to remove higher frequency components. Then, it is processed by the instantaneous frequency to obtain the fundamental frequency estimates. The symmetric higher order energy operator has been used as an indicator to determine the voiced/unvoiced speech. The fundamental frequency estimates are further processed by a moving average filter to obtain the monotonically changed estimates. The obtained fundamental frequency estimates have been compared with the spectrogram of the speech to confirm its accuracy.

• 한국소음진동공학회논문집
• /
• 제16권3호
• /
• pp.255-262
• /
• 2006

Frequency spectrum using the conventional Fourier analysis gives adequate information about the dynamic characteristics of the slider air bearing for the linear and stationary cases. The intermittent contacts for the extremely low flying height, however, generate nonlinear and nonstationary vibration at the instant of contact. Nonlinear dynamic model should be developed to simulate the impulse response of the air bearing during slider-disk contact. Time-frequency analysis is widely used to investigate the nonstationary signal. Several time-frequency analysis methods are employed and compared for the slider vibration signal caused by the impact against an artificially induced scratch on the disk. The representative Wigner-Ville distribution leads to the severe interference problem by cross terms even though it gives good resolution both in time and frequency. The smoothing process improves the interference problem at the expense of resolution. In order to get the results with good resolution and little interference, the reassignment method is proposed. Among others the reassigned Gabor spectrogram shows the best resolution and readability with negligible interference.

• 한국음향학회지
• /
• 제38권6호
• /
• pp.687-696
• /
• 2019

기존의 지진파 분류 특징은 강진에 초점이 맞추어져 있어서 미소지진과 같은 지진파는 다소 적합하지 않다. 본 연구에서는 강진과 더불어 미소지진, 인공지진, 잡음 분류에 적합한 특징 추출을 위해 주파수-시간 공간 내에서 히스토그램과 주성분 기반 특징 추출방법을 제안한다. 제안된 방법은 지진파의 주파수 관련 정보와 시간 관련 정보를 결합하는 방법을 적용한 히스토그램 기반 특징 추출방법과 주성분 기반 특징 추출방법을 이용하여 지진(강진, 미소지진, 인공지진)과 잡음, 미소지진과 잡음, 미소지진과 인공지진을 이진 분류한다. 2017년~2018년 최근 국내지진 자료와 분류 성능을 토대로 제안한 특징 추출방식의 효용성을 비교 평가한다.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2006년도 하계종합학술대회
• /
• pp.819-820
• /
• 2006

The purpose of this study is to develop an automatic algorithm to detect the arousal events. The proposed method is based on time-frequency analysis and the support vector machine classifier using single channel electroencephalogram. To extract features, first we computed 6 indices to find out the information of sleep states. Next powers of each of 4 frequency bands were computed using spectrogram of arousal region. And finally we computed variations of power of EEG frequency to detect arousals. The performance has been assessed using polysomnographic recordings of twenty patients with sleep apnea, snoring and excessive daytime sleepiness. We have shown that proposed method was effective for detecting the arousal events.

• Structural Monitoring and Maintenance
• /
• 제9권2호
• /
• pp.107-127
• /
• 2022

Structural health monitoring (SHM) has been related to damage identification with either operational loads or other environmental loading playing a significant complimentary role in terms of structural safety. In this study, a non-parametric method of time frequency analysis on the measurement is used to address the time-frequency representation for modal parameter estimation and system damage identification of structure. The method employs the wavelet decomposition of dynamic data by using the modified complex Morlet wavelet with variable central frequency (MCMW+VCF). Through detail discussion on the selection of model parameter in wavelet analysis, the method is applied to study the dynamic response of both steel structure and reinforced concrete frame under white noise excitation as well as earthquake excitation from shaking table test. Application of the method to building earthquake response measurement is also examined. It is shown that by using the spectrogram generated from MCMW+VCF method, with suitable selected model parameter, one can clearly identify the time-varying modal frequency of the reinforced concrete structure under earthquake excitation. Discussions on the advantages and disadvantages of the method through field experiments are also presented.

• Smart Structures and Systems
• /
• 제29권4호
• /
• pp.625-640
• /
• 2022

Maglev rail joints are vital components serving as connections between the adjacent F-type rail sections in maglev guideway. Damage to maglev rail joints such as bolt looseness may result in rough suspension gap fluctuation, failure of suspension control, and even sudden clash between the electromagnets and F-type rail. The condition monitoring of maglev rail joints is therefore highly desirable to maintain safe operation of maglev. In this connection, an online damage detection approach based on three-dimensional (3D) convolutional neural network (CNN) and time-frequency characterization is developed for simultaneous detection of multiple damage of maglev rail joints in this paper. The training and testing data used for condition evaluation of maglev rail joints consist of two months of acceleration recordings, which were acquired in-situ from different rail joints by an integrated online monitoring system during a maglev train running on a test line. Short-time Fourier transform (STFT) method is applied to transform the raw monitoring data into time-frequency spectrograms (TFS). Three CNN architectures, i.e., small-sized CNN (S-CNN), middle-sized CNN (M-CNN), and large-sized CNN (L-CNN), are configured for trial calculation and the M-CNN model with excellent prediction accuracy and high computational efficiency is finally optioned for multiple damage detection of maglev rail joints. Results show that the rail joints in three different conditions (bolt-looseness-caused rail step, misalignment-caused lateral dislocation, and normal condition) are successfully identified by the proposed approach, even when using data collected from rail joints from which no data were used in the CNN training. The capability of the proposed method is further examined by using the data collected after the loosed bolts have been replaced. In addition, by comparison with the results of CNN using frequency spectrum and traditional neural network using TFS, the proposed TFS-CNN framework is proven more accurate and robust for multiple damage detection of maglev rail joints.

• 대한전기학회논문지:시스템및제어부문D
• /
• 제55권5호
• /
• pp.240-247
• /
• 2006

Frequent arousals during sleep degrade the quality of sleep and result in sleep fragmentation. Visual inspection of physiological signals to detect the arousal events is cumbersome and time-consuming work. The purpose of this study is to develop an automatic algorithm to detect the arousal events. The proposed method is based on time-frequency analysis and the support vector machine classifier using single channel electroencephalogram (EEG). To extract features, first we computed 6 indices to find out the informations of a subject's sleep states. Next powers of each of 4 frequency bands were computed using spectrogram of arousal region. And finally we computed variations of power of EEG frequency to detect arousals. The performance has been assessed using polysomnographic (PSG) recordings of twenty patients with sleep apnea, snoring and excessive daytime sleepiness (EDS). We could obtain sensitivity of 79.65%, specificity of 89.52% for the data sets. We have shown that proposed method was effective for detecting the arousal events.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권2호
• /
• pp.348-369
• /
• 2024

With the advancement of Industry 4.0 and Industrial Internet of Things (IIoT), manufacturing increasingly seeks automation and intelligence. Temperature and vibration monitoring are essential for machinery health. Traditional abnormal state detection methodologies often overlook the intricate frequency characteristics inherent in vibration time series and are susceptible to erroneously reconstructing temperature abnormalities due to the highly similar waveforms. To address these limitations, we introduce synergistic, end-to-end, unsupervised Frequency-Time Domain Memory-Enhanced Autoencoders (FTD-MAE) capable of identifying abnormalities in both temperature and vibration datasets. This model is adept at accommodating time series with variable frequency complexities and mitigates the risk of overgeneralization. Initially, the frequency domain encoder processes the spectrogram generated through Short-Time Fourier Transform (STFT), while the time domain encoder interprets the raw time series. This results in two disparate sets of latent representations. Subsequently, these are subjected to a memory mechanism and a limiting function, which numerically constrain each memory term. These processed terms are then amalgamated to create two unified, novel representations that the decoder leverages to produce reconstructed samples. Furthermore, the model employs Spectral Entropy to dynamically assess the frequency complexity of the time series, which, in turn, calibrates the weightage attributed to the loss functions of the individual branches, thereby generating definitive abnormal scores. Through extensive experiments, FTD-MAE achieved an average ACC and F1 of 0.9826 and 0.9808 on the CMHS and CWRU datasets, respectively. Compared to the best representative model, the ACC increased by 0.2114 and the F1 by 0.1876.

• 정보처리학회 논문지
• /
• 제13권6호
• /
• pp.284-290
• /
• 2024

음성감정인식(Speech Emotion Recognition, SER)은 사용자의 목소리에서 나타나는 떨림, 어조, 크기 등의 음성 패턴 분석을 통하여 감정 상태를 판단하는 기술이다. 하지만, 기존의 음성 감정인식 연구는 구현된 시나리오를 이용하여 제한된 환경 내에서 숙련된 연기자를 대상으로 기록된 음성인 구현발화를 중심의 연구로 그 결과 또한 높은 성능을 얻을 수 있지만, 이에 반해 자유발화 감정인식은 일상생활에서 통제되지 않는 환경에서 이루어지기 때문에 기존 구현발화보다 현저히 낮은 성능을 보여주고 있다. 본 논문에서는 일상적 자유발화 음성을 활용하여 감정인식을 진행하고, 그 성능을 향상하고자 한다. 성능평가를 위하여 AI Hub에서 제공되는 한국인 자유발화 대화 음성데이터를 사용하였으며, 딥러닝 학습을 위하여 1차원의 음성신호를 시간-주파수가 포함된 2차원의 스펙트로그램(Spectrogram)로 이미지 변환을 진행하였다. 생성된 이미지는 CNN기반 전이학습 신경망 모델인 VGG (Visual Geometry Group) 로 학습하였고, 그 결과 7개 감정(기쁨, 사랑스러움, 화남, 두려움, 슬픔, 중립, 놀람)에 대해서 성인 83.5%, 청소년 73.0%의 감정인식 성능을 확인하였다. 본 연구를 통하여, 기존의 구현발화기반 감정인식 성능과 비교하면, 낮은 성능이지만, 자유발화 감정표현에 대한 정량화할 수 있는 음성적 특징을 규정하기 어려움에도 불구하고, 일상생활에서 이루어진 대화를 기반으로 감정인식을 진행한 점에서 의의를 두고자 한다.